Predictive modeling of phishing websites: a comprehensive analysis of novel traits using extra tree classification and optimized hybridization tactics

Phishing is more and more becoming a standard online threat, mainly due to social networking, web technology, and mobile innovations. Phishing taxonomies have mainly targeted core phishing processes, not the newly developed attack tactics, target environments, or defenses against emerging phishing types. Phishing poses a serious problem in the field of computer science, and more so in e-commerce and e-banking, which are responsible for the count of online transactions with payments. Sophisticated phishing page evolution requires adaptive countermeasures. Relying on machine learning (ML) to provide dynamic defence against these cunning baits. ML schemes circumvent static blacklists and instead identify new as well as repeated phishing attempts at an early rate by a look at traits like URLs, content, and user behavior. However, an ongoing evolutionary element enters the picture with phishers continuously adjusting their modus operandi. The battle is to be a step ahead, and ML's adaptability has shown its true value in this endless battle of the digital world. Extra tree classification (ETC) was used in this study to detect such phishing websites. Optimization method was utilized in this study, using 2 optimizers, Chaos gaming optimization (CGO) and tasmanian devil optimization (TDO), in the original model to increase its performance. Compared to ETC and ETCG, the ETTD model functions very well, and it is the best model in this paper. 2 percent makes up the difference between the ETTD model and the ETCG model, while about 6 percent makes the difference between the ETTD model and the ETC model.